Sell a private GitHub, GitLab or Bitbucket repo to AI labs

A private repository on GitHub, GitLab or Bitbucket can be licensed to AI labs through Pangea, which pays $100 per million tokens of source and $40 per million tokens of git history and leaves you the code. The license is non-exclusive, so you keep shipping it. You grant read-only access to the host; which host it is makes no difference to the price. The worth calculator connects to all three and shows what the source would earn before you talk to anyone.

Private since day one, yours to license, and not a fork

Pangea’s page lists what it won’t take:

• Code that was ever public, even if the repository is private now.
• Forks, because the upstream code belongs to someone else.
• Client work, since an agency can’t license what it built for a client.
• Code with confidential data or conflicting third-party rights, such as a co-founder who never signed an IP assignment.

Pangea scrubs secrets, credentials and personal data from its own copy before a lab sees it, so a stray key in the history isn’t a problem. Ownership is: Pangea’s agreement makes you responsible for owning what you submit, with no cap.

Individuals go through Pangea; companies have more routes

Pangea takes repositories from anyone who owns them. A company with a codebase and an operating record has four other listed routes, each of which takes the code as part of a bigger record:

• Grepped takes codebases with their commits, reviews and issues from GitHub or GitLab, alongside a company’s other records.
• Miro Advisory takes private codebases with commits, issues, pull requests, reviews and tests, from companies of roughly 20 to 700 people, and publishes an indicative $10K to $1M+ per codebase.
• License My Data takes shared drives, code and design files from companies of 30 or more people, or 50 at peak for a closed company.
• Turing buys a company’s whole operating record in one deal, source code included, for a single payout of up to $1M.

The code fit check covers Pangea. A company should run the company check as well.

GitHub: connect the app, keep the branches

The calculator’s “Connect your code” signs in with a GitHub App that reads the repositories you own, their sizes and languages, and the commit count. Pangea’s own process asks for the same read-only access after you sign.

History on GitHub lives in branches and tags. A squash merge keeps one commit per pull request, and the versions that lived only on a deleted branch don’t count, because a clone never brings them along. If you still have long-lived feature branches, leave them in place.

GitLab: gitlab.com projects, with statistics

Connect reads your gitlab.com projects with their statistics and commit counts. Self-hosted GitLab isn’t covered by the sign-in, so for a repository on your own server, pick a local clone as a folder in the calculator or run pre.dev’s script on it. Pangea accepts GitLab access the same way as GitHub.

Bitbucket: one workspace at a time

Connect reads one Bitbucket workspace and pages through its commits up to 10,000, which shows as “10,000+” when the host confirms there are more. Pangea accepts Bitbucket access too.

Moving a repository between hosts: mirror it

If the code is on a host Pangea doesn’t connect to, or you want it under an account you control before you sign, move it with a mirror clone so every branch, tag and commit comes along:

git clone --mirror git@old-host:you/repo.git
cd repo.git
git push --mirror git@github.com:you/repo.git

A plain copy of the working folder, or a clone made with --depth, has no history to count. If pre.dev’s script shows almost no history for a repository you know is old, that is usually why.

The sale: estimate, sign, grade, license, pay

1. Request access on Pangea’s page and connect Git. Pangea returns an estimate.
2. Review and sign the agreement, then grant read-only access and choose the repositories.
3. Pangea scrubs and grades a separate copy, in a few hours to four weeks.
4. Accepted repositories go to labs in batches. Licensing typically takes 2 to 16 weeks and is not guaranteed.
5. You’re paid within 15 days of the buyer’s payment, once your tax and payout details are complete.

You can withdraw a repository before it goes to labs, and remove it from future opportunities on 30 days’ notice after that. Pangea’s rates can change on 30 days’ notice, never retroactively.

Keep the full history, proof of privacy and the IP paperwork

• Every branch and tag, unsquashed and unpruned.
• Proof the repository was never public: the host’s settings history, or your own records.
• IP assignments from co-founders and contractors, and any contract with a client or acquirer that touches the code.
• Read-only access ready to grant from an account the owner controls.